#include <sys/resource.h>
#include <sys/wait.h>
+void aofUpdateCurrentSize(void);
+
/* Called when the user switches from "appendonly yes" to "appendonly no"
* at runtime using the CONFIG command. */
void stopAppendOnly(void) {
server.appendseldb = -1;
server.appendonly = 0;
/* rewrite operation in progress? kill it, wait child exit */
- if (server.bgsavechildpid != -1) {
+ if (server.bgrewritechildpid != -1) {
int statloc;
- if (kill(server.bgsavechildpid,SIGKILL) != -1)
+ if (kill(server.bgrewritechildpid,SIGKILL) != -1)
wait3(&statloc,0,NULL);
/* reset the buffer accumulating changes while the child saves */
sdsfree(server.bgrewritebuf);
server.bgrewritebuf = sdsempty();
- server.bgsavechildpid = -1;
+ server.bgrewritechildpid = -1;
}
}
* buffer and write it on disk using this function just before entering
* the event loop again. */
void flushAppendOnlyFile(void) {
- time_t now;
ssize_t nwritten;
if (sdslen(server.aofbuf) == 0) return;
}
sdsfree(server.aofbuf);
server.aofbuf = sdsempty();
+ server.appendonly_current_size += nwritten;
- /* Don't Fsync if no-appendfsync-on-rewrite is set to yes and we have
- * childs performing heavy I/O on disk. */
+ /* Don't fsync if no-appendfsync-on-rewrite is set to yes and there are
+ * children doing I/O in the background. */
if (server.no_appendfsync_on_rewrite &&
(server.bgrewritechildpid != -1 || server.bgsavechildpid != -1))
return;
- /* Fsync if needed */
- now = time(NULL);
+
+ /* Perform the fsync if needed. */
if (server.appendfsync == APPENDFSYNC_ALWAYS ||
(server.appendfsync == APPENDFSYNC_EVERYSEC &&
- now-server.lastfsync > 1))
+ server.unixtime > server.lastfsync))
{
/* aof_fsync is defined as fdatasync() for Linux in order to avoid
* flushing metadata. */
aof_fsync(server.appendfd); /* Let's try to get this data on the disk */
- server.lastfsync = now;
+ server.lastfsync = server.unixtime;
}
}
FILE *fp = fopen(filename,"r");
struct redis_stat sb;
int appendonly = server.appendonly;
+ long loops = 0;
- if (redis_fstat(fileno(fp),&sb) != -1 && sb.st_size == 0)
+ if (fp && redis_fstat(fileno(fp),&sb) != -1 && sb.st_size == 0) {
+ server.appendonly_current_size = 0;
+ fclose(fp);
return REDIS_ERR;
+ }
if (fp == NULL) {
redisLog(REDIS_WARNING,"Fatal error: can't open the append log file for reading: %s",strerror(errno));
server.appendonly = 0;
fakeClient = createFakeClient();
+ startLoading(fp);
+
while(1) {
int argc, j;
unsigned long len;
char buf[128];
sds argsds;
struct redisCommand *cmd;
- int force_swapout;
+
+ /* Serve the clients from time to time */
+ if (!(loops++ % 1000)) {
+ loadingProgress(ftello(fp));
+ aeProcessEvents(server.el, AE_FILE_EVENTS|AE_DONT_WAIT);
+ }
if (fgets(buf,sizeof(buf),fp) == NULL) {
if (feof(fp))
redisLog(REDIS_WARNING,"Unknown command '%s' reading the append only file", argv[0]->ptr);
exit(1);
}
- /* Try object encoding */
- if (cmd->flags & REDIS_CMD_BULK)
- argv[argc-1] = tryObjectEncoding(argv[argc-1]);
/* Run the command in the context of a fake client */
fakeClient->argc = argc;
fakeClient->argv = argv;
/* The fake client should not have a reply */
redisAssert(fakeClient->bufpos == 0 && listLength(fakeClient->reply) == 0);
-
- /* Clean up, ready for the next command */
- for (j = 0; j < argc; j++) decrRefCount(argv[j]);
- zfree(argv);
-
- /* Handle swapping while loading big datasets when VM is on */
- force_swapout = 0;
- if ((zmalloc_used_memory() - server.vm_max_memory) > 1024*1024*32)
- force_swapout = 1;
-
- if (server.vm_enabled && force_swapout) {
- while (zmalloc_used_memory() > server.vm_max_memory) {
- if (vmSwapOneObjectBlocking() == REDIS_ERR) break;
- }
- }
+ /* The fake client should never get blocked */
+ redisAssert((fakeClient->flags & REDIS_BLOCKED) == 0);
+
+ /* Clean up. Command code may have changed argv/argc so we use the
+ * argv/argc of the client instead of the local variables. */
+ for (j = 0; j < fakeClient->argc; j++)
+ decrRefCount(fakeClient->argv[j]);
+ zfree(fakeClient->argv);
}
/* This point can only be reached when EOF is reached without errors.
fclose(fp);
freeFakeClient(fakeClient);
server.appendonly = appendonly;
+ stopLoading();
+ aofUpdateCurrentSize();
+ server.auto_aofrewrite_base_size = server.appendonly_current_size;
return REDIS_OK;
readerr:
redisDb *db = server.db+j;
dict *d = db->dict;
if (dictSize(d) == 0) continue;
- di = dictGetIterator(d);
+ di = dictGetSafeIterator(d);
if (!di) {
fclose(fp);
return REDIS_ERR;
/* Iterate this DB writing every entry */
while((de = dictNext(di)) != NULL) {
- sds keystr = dictGetEntryKey(de);
+ sds keystr;
robj key, *o;
time_t expiretime;
- int swapped;
keystr = dictGetEntryKey(de);
o = dictGetEntryVal(de);
initStaticStringObject(key,keystr);
- /* If the value for this key is swapped, load a preview in memory.
- * We use a "swapped" flag to remember if we need to free the
- * value object instead to just increment the ref count anyway
- * in order to avoid copy-on-write of pages if we are forked() */
- if (!server.vm_enabled || o->storage == REDIS_VM_MEMORY ||
- o->storage == REDIS_VM_SWAPPING) {
- swapped = 0;
- } else {
- o = vmPreviewObject(o);
- swapped = 1;
- }
+
expiretime = getExpire(db,&key);
/* Save the key and associated value */
}
} else if (o->type == REDIS_ZSET) {
/* Emit the ZADDs needed to rebuild the sorted set */
- zset *zs = o->ptr;
- dictIterator *di = dictGetIterator(zs->dict);
- dictEntry *de;
-
- while((de = dictNext(di)) != NULL) {
- char cmd[]="*4\r\n$4\r\nZADD\r\n";
- robj *eleobj = dictGetEntryKey(de);
- double *score = dictGetEntryVal(de);
-
- if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
- if (fwriteBulkObject(fp,&key) == 0) goto werr;
- if (fwriteBulkDouble(fp,*score) == 0) goto werr;
- if (fwriteBulkObject(fp,eleobj) == 0) goto werr;
+ char cmd[]="*4\r\n$4\r\nZADD\r\n";
+
+ if (o->encoding == REDIS_ENCODING_ZIPLIST) {
+ unsigned char *zl = o->ptr;
+ unsigned char *eptr, *sptr;
+ unsigned char *vstr;
+ unsigned int vlen;
+ long long vll;
+ double score;
+
+ eptr = ziplistIndex(zl,0);
+ redisAssert(eptr != NULL);
+ sptr = ziplistNext(zl,eptr);
+ redisAssert(sptr != NULL);
+
+ while (eptr != NULL) {
+ redisAssert(ziplistGet(eptr,&vstr,&vlen,&vll));
+ score = zzlGetScore(sptr);
+
+ if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
+ if (fwriteBulkObject(fp,&key) == 0) goto werr;
+ if (fwriteBulkDouble(fp,score) == 0) goto werr;
+ if (vstr != NULL) {
+ if (fwriteBulkString(fp,(char*)vstr,vlen) == 0)
+ goto werr;
+ } else {
+ if (fwriteBulkLongLong(fp,vll) == 0)
+ goto werr;
+ }
+ zzlNext(zl,&eptr,&sptr);
+ }
+ } else if (o->encoding == REDIS_ENCODING_SKIPLIST) {
+ zset *zs = o->ptr;
+ dictIterator *di = dictGetIterator(zs->dict);
+ dictEntry *de;
+
+ while((de = dictNext(di)) != NULL) {
+ robj *eleobj = dictGetEntryKey(de);
+ double *score = dictGetEntryVal(de);
+
+ if (fwrite(cmd,sizeof(cmd)-1,1,fp) == 0) goto werr;
+ if (fwriteBulkObject(fp,&key) == 0) goto werr;
+ if (fwriteBulkDouble(fp,*score) == 0) goto werr;
+ if (fwriteBulkObject(fp,eleobj) == 0) goto werr;
+ }
+ dictReleaseIterator(di);
+ } else {
+ redisPanic("Unknown sorted set encoding");
}
- dictReleaseIterator(di);
} else if (o->type == REDIS_HASH) {
char cmd[]="*4\r\n$4\r\nHSET\r\n";
if (fwriteBulkObject(fp,&key) == 0) goto werr;
if (fwriteBulkLongLong(fp,expiretime) == 0) goto werr;
}
- if (swapped) decrRefCount(o);
}
dictReleaseIterator(di);
}
*/
int rewriteAppendOnlyFileBackground(void) {
pid_t childpid;
+ long long start;
if (server.bgrewritechildpid != -1) return REDIS_ERR;
- if (server.vm_enabled) waitEmptyIOJobsQueue();
+ start = ustime();
if ((childpid = fork()) == 0) {
- /* Child */
char tmpfile[256];
- if (server.vm_enabled) vmReopenSwapFile();
- close(server.fd);
+ /* Child */
+ if (server.ipfd > 0) close(server.ipfd);
+ if (server.sofd > 0) close(server.sofd);
snprintf(tmpfile,256,"temp-rewriteaof-bg-%d.aof", (int) getpid());
if (rewriteAppendOnlyFile(tmpfile) == REDIS_OK) {
_exit(0);
}
} else {
/* Parent */
+ server.stat_fork_time = ustime()-start;
if (childpid == -1) {
redisLog(REDIS_WARNING,
"Can't rewrite append only file in background: fork: %s",
void bgrewriteaofCommand(redisClient *c) {
if (server.bgrewritechildpid != -1) {
addReplyError(c,"Background append only file rewriting already in progress");
- return;
- }
- if (rewriteAppendOnlyFileBackground() == REDIS_OK) {
+ } else if (server.bgsavechildpid != -1) {
+ server.aofrewrite_scheduled = 1;
+ addReplyStatus(c,"Background append only file rewriting scheduled");
+ } else if (rewriteAppendOnlyFileBackground() == REDIS_OK) {
addReplyStatus(c,"Background append only file rewriting started");
} else {
addReply(c,shared.err);
unlink(tmpfile);
}
+/* Update the server.appendonly_current_size filed explicitly using stat(2)
+ * to check the size of the file. This is useful after a rewrite or after
+ * a restart, normally the size is updated just adding the write length
+ * to the current lenght, that is much faster. */
+void aofUpdateCurrentSize(void) {
+ struct redis_stat sb;
+
+ if (redis_fstat(server.appendfd,&sb) == -1) {
+ redisLog(REDIS_WARNING,"Unable to check the AOF length: %s",
+ strerror(errno));
+ } else {
+ server.appendonly_current_size = sb.st_size;
+ }
+}
+
/* A background append only file rewriting (BGREWRITEAOF) terminated its work.
* Handle this. */
-void backgroundRewriteDoneHandler(int statloc) {
- int exitcode = WEXITSTATUS(statloc);
- int bysignal = WIFSIGNALED(statloc);
-
+void backgroundRewriteDoneHandler(int exitcode, int bysignal) {
if (!bysignal && exitcode == 0) {
int fd;
char tmpfile[256];
if (server.appendfsync != APPENDFSYNC_NO) aof_fsync(fd);
server.appendseldb = -1; /* Make sure it will issue SELECT */
redisLog(REDIS_NOTICE,"The new append only file was selected for future appends.");
+ aofUpdateCurrentSize();
+ server.auto_aofrewrite_base_size = server.appendonly_current_size;
} else {
/* If append only is disabled we just generate a dump in this
* format. Why not? */
} else {
redisLog(REDIS_WARNING,
"Background append only file rewriting terminated by signal %d",
- WTERMSIG(statloc));
+ bysignal);
}
cleanup:
sdsfree(server.bgrewritebuf);
projects / redis.git / blobdiff